Sushmitha Ramesh

About

Cloud/ML engineer with production deployment experience on Azure (Dockerized models, managed APIs, data pipelines) who has repeatedly stabilized unreliable systems—e.g., taking an API-driven analytics pipeline from ~60% to 98% reliability and an Azure ML service from ~80% to 97% by addressing rate limits, container memory, and gateway timeouts. Also built an explainable contract-risk model for entertainment bookings (Transformers + SHAP) and integrated it into a legacy booking system via a Flask REST API, plus prior IoT work at Nissan processing CAN bus sensor streams for diagnostics/anomaly insights.

Key Strengths

• Improved production reliability from ~60% to 98% by adding logging, exponential backoff retries, and schema validation for flaky third-party APIs
• Systematic multi-layer debugging across model, container resources, and networking/API gateway timeouts (raised reliability ~80% to 97%)
• Optimized ML feature extraction via chunked document processing to reduce memory spikes and latency under concurrent load (10–15 concurrent requests)
• Bridged legacy systems by wrapping an ML model with a lightweight Flask REST API and deploying on Azure
• Built stakeholder trust and adoption using SHAP explainability and close collaboration with booking agents (pilot, daily monitoring, iterative fixes)
• Hands-on IoT/automotive telemetry processing using CAN bus data with real-time Python pipelines

Experience

Education

Awards

• Government research grant and best project (Karnataka, India) - Drowsiness Driver Detection System

Skills

Languages